Are imports handled differently when packaging code into an AWS Lambda layer?

When you package code directly in a Lambda function, the runtime looks for modules in the function’s deployment package root or in the /var/task directory. A Lambda layer, however, is a separate ZIP that is mounted at /opt during execution. Therefore, any modules you place in a layer must be under a subdirectory that matches the runtime’s expected path (e.g., python/lib/python3.9/site-packages for Python). Imports in your function code then reference those modules as if they were in the function’s own directory. This separation allows you to share common libraries across multiple functions, keep deployment packages small, and update libraries independently of the function code. It also means you must ensure the layer’s runtime matches the function’s runtime and that you reference the layer’s ARN correctly in the function configuration.